+ "--task", "classify"

Author: noooop

tests/entrypoints/openai/test_classification.py

@@ -18,6 +18,8 @@ def server():
         "--enforce-eager",
         "--max-model-len",
         "512",
+        "--task",
+        "classify",
         "--dtype",
         DTYPE,
     ]

tests/models/language/pooling/test_classification.py

@@ -37,7 +37,8 @@ def test_models(
         # switch to use ROCm CK FA backend
         monkeypatch.setenv("VLLM_USE_TRITON_FLASH_ATTN", "False")
 
-    with vllm_runner(model, max_model_len=512, dtype=dtype) as vllm_model:
+    with vllm_runner(model, max_model_len=512, dtype=dtype,
+                     task="classify") as vllm_model:
         vllm_outputs = vllm_model.classify(example_prompts)
 
     with hf_runner(model,